Debug AI agent tool calls with OpenRouter, calculator and webhook

Quick overview This workflow gives a support agent a Calculator tool and a simulated order lookup tool, so you can compare a correct calculation, a found order and an empty lookup that still reports success. It is Exercise 4 from the n8n Production AI Playbook on Audit & Trace.

How it works Receives a POST webhook request with a natural-language query. Support Agent, an AI Agent using an OpenRouter chat model, decides whether to call a tool. Calculator Tool handles arithmetic questions such as 47 times 89. Order Lookup Tool returns a simulated order record, or an empty string when the order does not exist. Respond returns the agent's answer to the webhook caller.

Setup Add an OpenRouter API credential to OpenRouter - Agent. If you change models, choose one that supports tool calling. Optionally enable OpenTelemetry tracing in Settings > OpenTelemetry to see a span for each tool call. Select Execute workflow and POST {"query":"What is 47 times 89?"} to the Webhook Test URL. Repeat with {"query":"How much was order ORD-1001? Use the order lookup."} and {"query":"How much was order ORD-8888? Tell the customer the amount."}. Compare the tool results in the agent logs with each reply to find the empty lookup.

Requirements n8n Cloud or a self-hosted n8n instance An OpenRouter API key for a model with tool-calling support

Customization Have Order Lookup Tool return a clear message such as "No order found" instead of an empty string, so the agent can tell the customer the order is missing.

Additional info Order data is synthetic.